[AstroDave]

Two month’s ago (!), inspired by this posting (Shutter speed tester - PentaxForums.com) about measuring shutter speeds for a film camera, I put together a simple system to make such measurements, using my Pentax ME Super A as the device under test (DUT in electronics jargon).

(Aside - the photoplug tester used by forum member Vendee is almost certainly using an integrating detector with a modest time constant - I could duplicate those swoopy light curves when I used my photodetector in an integrator circuit.)

Here’s a shot of the setup: I configured a cheap green laser pointer module from Ebay to shoot through the ME Super with its film back open and on to a high-speed photodiode light detector. The photodiode response (scope trace goes high when light shines on the diode) was displayed on a storage oscilloscope, from which I could make screen dump copies of the displayed light curves. The overall response time of the system is at the microsecond (millionth of second) level.




The scope also has a pair of cursors that can be placed over the time response trace. I set those at the half-power levels at the start and end of the light pulse that came through the shutter and read directly the time interval. This interval is how long the shutter is open. Typically, I made 2 or 3 measurements of the shutter time. The values given here are the average of those measurements.

For longer durations, more than a 1/30 second or so, the rise and fall of the light curve is very sharp at the time resolution of the scope when showing the entire open trace. For shorter shutter speeds, the rise/fall times are resolved, and at the very shortest times (1/1000 and 1/2000 seconds), the shutter is barely “open” for the given duration of the exposure.

Here are the results in tabular form, with the nominal shutter speed (i.e. the speed shown in the camera display - the ME Super shows the shutter speed in the viewfinder display), the nominal speed given in milliseconds, the measured shutter speed in milliseconds, and the ratio between them, in the sense of measured divided by nominal.

Nominal	Nominal	Measured	Ratio
(sec)	(millisec)	(millisec)
1/2000	0.50	0.44	0.88
1/1000	1	0.87	0.87
1/500	2	1.82	0.91
1/250	4	3.82	0.96
1/125	8	9.80	1.22
1/60	16.7	15.4	0.92
1/30	33.3	30.6	0.92
1/15	66.7	60	0.90
1/8	125	119	0.95
1/4	250	239	0.96
1/2	500	475	0.95
1	1000	961	0.96
2	2000	1916	0.96
4	4000	3820	0.96

And here are a few of the scope screen dumps.

For a quarter of a second (the vertical purple lines are the time measurement cursors) the edges are nice and sharp:




Originally, I wondered why there were those wiggles at the beginning of the trace (they were on most of the traces). Then, I realized that the camera was sitting on the board to which the rest of the system was mounted. The shutter slap was bouncing things around, in particular the detector diode with respect to the laser beam. When I picked the camera up off the board before I fired it, things were much better:




At short shutter speeds, the traces are not quite so distinct. Here’s what you see for 1/500 of a second. The open and close parts of the trace take place over a few hundred microseconds (the time divisions in this image are 200 microseconds):




And, at 1/2000 second, all you get is a bump, with about 400 microsecond rise and fall times:




Here’s a plot of the results. The speeds are consistently shorter than nominal, by about 5-10% (around a tenth of a stop) and even shorter at the two fastest speeds, except for the 1/125 speed. This is also the speed that is set for flash exposure. I measured the 1/125 setting for both the normal speed setting as well as the X (flash) position on the dial. In both cases, the measured speeds were comparable and longer than nominal. I wonder if Pentax gave themselves a bit of leeway here to make sure that for flash exposures, the shutter really would be open “long enough” for the flash.




Of course, these values apply only to my vintage ME Super (I'm the original owner), which has never been CLA’d and was last used seriously for a film class I took back in 2007 or so.


The rise and fall times tell us something about how fast the shutter curtains are moving. An a priori estimate of the curtain speed can be made by noting that the flash duration setting corresponds to a speed of 1/125 second. At shorter speeds, the second curtain starts its motion before the first curtain has had a chance to fully traverse the 24 mm vertical frame. So, we estimate that it takes around 1/125 second for a shutter curtain to move 24 mm. This corresponds to a curtain speed of

24mm/(1/125 sec) = 3000 mm/ sec (pretty fast!)

How should this affect my scope traces? Well, my green laser pointer produces a beam of light which is slightly bigger than 1 mm in diameter (with an unknown, but probably not too far from Gaussian, profile). If I naively assume a uniform dot 1 mm across, it should take a shutter curtain about 1 mm/3000 mm/sec = 333 microseconds to slice across the beam. This is remarkably close to what I see.

This image shows zoomed-in views of both the open and close light traces for a moderate shutter speed (1/60 second). (The close trace has been reversed and overlaid on the rise trace, with an eyeball best fit.) The whole process takes around 400 microseconds - very consistent with the estimate above for how long it should take a curtain to move across the laser beam. These timings are also consistent with the “bump” shown above for the 1/2000 second measurement.

[MarkJerling]

Wow, impressive work Dave!

[LesDMess]

Two month’s ago (!), inspired by this posting (Shutter speed tester - PentaxForums.com) about measuring shutter speeds for a film camera, I put together a simple system to make such measurements, using my Pentax ME Super A as the device under test (DUT in electronics jargon).

Thanks for sharing!
Now can you boil it down to an ap . . .

[swanlefitte]

Nice work. You mentions 2-3 times per setting. I assume they were close or you would have mentioned the variation. Was there more variation at the 1/125 speed? As an outlier i wonder. Your idea it is headway for the max synch speed has merit.

Now go back in time and do it to the new camera. Will the fast speeds become more linear? I wonder what the tolerances were to pass inspection. Seems great to me new and after all these years.
Is the light meter also around 95%? Does it spike at 1/125?
Definately great tests because it makes me think of more.

[Wojciech]

Thanks for posting. I like your setup very much and I also like your critical discussion of the data.
When I was looking at your data it become apparent to me the (excluding the 1/125 sec data point) that the shutter is very accurate from 4 to 1/8 sec after which it seems to be "faster" then expected from the trend on the slow shutter values.
I will make a very unscientific statement here: I read somewhere (don't remember where, but it made perfect sense to me) that the actual shutter speeds above 1/8 are approximated in a weird way.
Now to be more scientific here is the explanation:
For each step in EV, each next step on the shutter should be exactly 2 times faster. So the next steps after 1/8 should idealy be 1/32, then 1/64, then 1/128 etc. Regarding the fact that these 1/2n values should give the most accurate relative exposure and it also seems to be simpler to calibrate the shutter in such way that each step is twice faster rather than calibrating it to a specific value, I would assume that the nominal times in your table are not the "real" nominal times.

If one calculates the expected actual times the ratios become consistent with your data between 1/8 and 4 sec. (Sorry for the poor look of the table. I don't know how to insert a table in HTML)
(Thanks for the tip on how to add the table AstroDave)

Nominal	Actual	Actual	Measured	Ratio
s	s	ms	ms
1/2000	1/2048	0,49	0,44	0,90
1/1000	1/1024	0,98	0,87	0,89
1/500	1/512	1,95	1,82	0,93
1/250	1/256	3,91	3,82	0,98
1/125	1/128	7,81	9,80	1,25
1/60	1/64	15,63	15,40	0,99
1/30	1/32	31,25	30,60	0,98
1/15	1/16	62,50	60,00	0,96

I like your idea that at 1/125 the shutter has to be open long enough to allow the exposure of the full frame during flash. Seems very plausible. The other factor might be that (if I am not wrong) the 1/125 shutter was mechanicaly operated on ME while all the other were electronically operated.

Thanks again for an interesting post.
Wojciech

[Not a Number]

Excellent job. Impress rig and similar to what repair shops use.

Now compare the results to one of those apps that use the sound of the shutter to measure the speed.

[AstroDave]

I will make a very unscientific statement here: I read somewhere (don't remember where, but it made perfect sense to me) that the actual shutter speeds above 1/8 are approximated in a weird way.
Now to be more scientific here is the explanation:
For each step in EV, each next step on the shutter should be exactly 2 times faster. So the next steps after 1/8 should idealy be 1/32, then 1/64, then 1/128 etc. Regarding the fact that these 1/2n values should give the most accurate relative exposure and it also seems to be simpler to calibrate the shutter in such way that each step is twice faster rather than calibrating it to a specific value, I would assume that the nominal times in your table are not the "real" nominal times.

Could be, and I even knew about this - see my post on "true" shutter speeds: True Shutter Speeds compared to Nominal Shutter Speeds - Actual Measurements - PentaxForums.com and references therein.

I don't know whether cameras of the ME Super vintage used electronic shutter control or not. Maybe somebody here can enlighten us on this. If they do, your suggestion makes a lot of sense, and the results do hang together better except for the 3 fastest speeds. But there is still the mystery of the 1/125 speed.

(Sorry for the poor look of the table. I don't know how to insert a table in HTML)

Take a look here: PentaxForums.com - BB Code List

I used the table code to stick my results in. It took a try or two to get the syntax correct. If you'd like to see the actual text I used, let me know and I can send you a PM or email.

Nice work. You mentions 2-3 times per setting. I assume they were close or you would have mentioned the variation. Was there more variation at the 1/125 speed? As an outlier i wonder. Your idea it is headway for the max synch speed has merit.

Thank you. The variations are indeed very small. At 1/125, there is a bit of systematic difference between the viewfinder value (values of 10.1, 10.0, and 10.1 milliseconds) and the X dial setting (9.38 and 9.40 milliseconds). If you take the overall scatter of these 5 values, the standard deviation is about 0.3 (only ~3%) and by far bigger than for any other of the speeds. For instance, at 1/8 second, I got 117, 120, 119, 119, 119, and 120 milliseconds. This is where I started and took a bunch of measurements to make sure I understood the system.

Now compare the results to one of those apps that use the sound of the shutter to measure the speed.

Now can you boil it down to an ap . . .

'Fraid not. Despite my technical prowess, I am a dinosaur when it comes to cell phones! I have a flip phone (which costs me about $20/year). I'm retired and we have no kids, so no need for instant/constant communications!!!!!

[Wojciech]

Could be, and I even knew about this - see my post on "true" shutter speeds: True Shutter Speeds compared to Nominal Shutter Speeds - Actual Measurements - PentaxForums.com and references therein.

I don't know whether cameras of the ME Super vintage used electronic shutter control or not. Maybe somebody here can enlighten us on this. If they do, your suggestion makes a lot of sense, and the results do hang together better except for the 3 fastest speeds. But there is still the mystery of the 1/125 speed.



Take a look here: PentaxForums.com - BB Code List

I used the table code to stick my results in. It took a try or two to get the syntax correct. If you'd like to see the actual text I used, let me know and I can send you a PM or email.

Thanks for the tip on the table. I fixed it now. Looks better.
I have never had a ME Super in my hand but Wikipedia says it has electric shutter and without batteries it only runs at 1/125 (or according to your measurements 1/102 s ) so I assume that at this speed it is purely mechanical.
I was not aware of your other work on shutter speeds. Thanks for referring to it. I'm impressed with your solution for the digital camera shutter speed measurement problem.

[AstroDave]

I have never had a ME Super in my hand but Wikipedia says it has electric shutter and without batteries it only runs at 1/125 (or according to your measurements 1/102 s ) so I assume that at this speed it is purely mechanical.
I was not aware of your other work on shutter speeds. Thanks for referring to it. I'm impressed with your solution for the digital camera shutter speed measurement problem.

I never thought to check Wikipedia (or the Forum review of the Super)! I did look in Wikipedia about shutters, but not specifically the ME Super (duh)! I think that explains a lot - especially running mechanically at (more-or-less) 1/125. I wouldn't be surprised if my measurements of the fastest speeds are somehow biased by my technique - taking the half-power levels on the rise and fall does not necessarily give the effective shutter time. I'd need to integrate the curve, being careful about amplitude calibration.

I was impressed with how accurate the speeds were for what I was thinking was a mechanical shutter that was ~40 years old! Now I know better. Some years back, I used a similar set up to measure an old (surely mechanical!) Topcon that my father had, and the values there were a lot farther from nominal.

I've improved my LED flasher system to a 256 LED array. I need to report those results (for K-1, K-3, Q7, and a couple of Canons: G15 and Rebel XS). The surprise there is that there is a very systematic deviation from nominal when the big Pentaxes are used for 1/3 shutter speed intervals, alternating between 3% high and 3% low between the "standard" values.

[swanlefitte]

I've improved my LED flasher system to a 256 LED array. I need to report those results (for K-1, K-3, Q7, and a couple of Canons: G15 and Rebel XS). The surprise there is that there is a very systematic deviation from nominal when the big Pentaxes are used for 1/3 shutter speed intervals, alternating between 3% high and 3% low between the "standard" values.

What do these models report as half steps? I notice my ks-2 uses f/4 f/4.5 f/5.6, not f/4.8. Third steps are as expected with f/4.5 and f/5.0.

[AstroDave]

What do these models report as half steps? I notice my ks-2 uses f/4 f/4.5 f/5.6, not f/4.8.

I checked - the K-1 and K-3 indeed report f/4.5 between f/4.0 and f/5.6 when in half-step mode. That is just plain WRONG!! WOW - Pentax get your act together. I hope others can check other models. At some point, I'll check to see if the actual exposures that result are equally wrong.

The Q7 and G15 have only 1/3 step mode. And, the Canon Rebel XS was borrowed from a friend, so I can't check it conveniently.

I thought it might have to do with what lens is attached (i.e. if the largest opening is close to 4.5), but I get the same results with the 100WR macro attached (minimum f-stop of 2.8). The 1/3 stops seem to be listed correctly.

[swanlefitte]

Thought it might be all pentax models. That makes these tests more interesting. Perhaps the shutter when on f/4.5 as a third vs as the half changes. The test can "shed some light" on this.

[peter.d]

I think that explains a lot - especially running mechanically at (more-or-less) 1/125.

Well, the 1/125 is purely mechanical however I believe the mode switch has to be set to 1/125 (not sure the exact marking, my ME has AUTO and 1/100) to fire without batteries.

My gut feeling tels me that in AUTO mode the shutter speed of 1/125 is still governed by electronics and the shutter solenoid. Might be wrong, though.

I wouldn't be surprised if my measurements of the fastest speeds are somehow biased by my technique - taking the half-power levels on the rise and fall does not necessarily give the effective shutter time.

I used the very same technique for measuring shutter speeds on my cameras. The faster shutter speeds have always been problematic due to a slow response time of the used light sensor. I got much better results when using IR LED+phototransistor from an old floppy drive. This combination showed much steeper rise curve.

In general, I would consider deviation of +/- 25% more than adequate for a film camera.

[AstroDave]

I used the very same technique for measuring shutter speeds on my cameras. The faster shutter speeds have always been problematic due to a slow response time of the used light sensor. I got much better results when using IR LED+phototransistor from an old floppy drive. This combination showed much steeper rise curve.

Guess I'm lazy - I just bought my photodiodes (QSD2030)! They were only 60 cents a piece - the postage is as much. They have VERY fast response time - 5 nanoseconds! (Depends on load resistance - I use a few kilohms which slows them down a bit)

"Floppy drive"? or CD-Rom or DVD - much more likely to contain phototransistors!

In general, I would consider deviation of +/- 25% more than adequate for a film camera.

Yeah - we are probably worrying way too much about this!

[swanlefitte]

In general, I would consider deviation of +/- 25% more than adequate for a film camera.

I would think if you did your own developing this would be much less an issue than if you sent them out.